GOPY AVAILABLE TO 00 DOES NOT PERMIT FULLY LEGIBLE PeOUCTIO D D C. A (.- Tit CANALYSIS OF CHEMICAL SMOKE RELEASES TO

Transcription

1 A (.- Tit CANALYSIS OF CHEMICAL SMOKE RELEASES TO. CHARACTERIZE STRATOSPHERIC / THERMOSPHERIC WIND FIELDS Sheldon B. Michaels Jeffrey S. Morris Otis Philbrick Information Design, Inc. Civil Air Terminal Bedford, Mass November 1976 Final Report for Period I January I November 1976 Approved for public release; distribution unlimited AIR FORCE GEOPHYSICS LABORATORY AIR FORCE SYSTEMS COMMAND UNITED STATES AIR FORCE HANSCOM AFB, MASSACHUSETTS D D C GOPY AVAILABLE TO 00 DOES NOT PERMIT FULLY LEGIBLE PeOUCTIO

2 Qualified requestors may obtain additional copies from the Defense Documentation Center. All others should apply to the National Technical Information Service.

3 SUCURITY CLAS116FICATION OF TNI% 044V (Men. b~eat.e ) "MOR DOCUMENTATION PAGE IA N1RCIN TOMPLETINGFOR BEFOR Inforaio Desgn Anc. /CVO NO. A A. WCORT UNI NUSERS IS. CONTOLW AN1D116 A0010ESS OFIC RAE Air FoeMCAAMK S Geophysic Laboratoryal-"OA MonitorAnto/i Qrsda CI Apreldor pbicelae ditibto imlixnited.±13 Stratosherickmaueet Civlerosperinmasueet Smdoe, relassacuet13 AiroFogrcmetropysc Laoaoy No7 Hnewo inter saciecmuetr program weree used tom AsuESe 14. MTeIN thermyoaph atfothrhill ADgtdfere n Canada Ofin Api195. PEUIYCAS osi&tonp DO ~ ~1#2 0 'MOV#IS OSOLETEUnclassified 0 1 S1URT C INILDISRIIIIIJ Xoer! STAT AIET (o Oho PlC.AS AU$l AGE 1-0~lttC~A SCM LE

4 PREFACE The authors wish to thank Dr. Antonio F. Quesada of AF(;L for his guidance and helpfulness during the development, testing, and operation of the film processing procedure. Io $a III i.-? JA1ft DDC (Con ID

5 INTRODUCTION S udies at AFGL on the moecxuinnt of clitntical simoke trails and puffs utilize photographs to define the trail or puff positions as a function of time. The chemical events are photographed at regular intervals with high resolution cameras from at least two separate sites. This report is concerned solely with the techniques used for recovering and digitizing the two-dimensional trail or puff image from the photographic films of the events. We report here work done on improving existing computer/microdensitometer techniques, on improvements which allow determination of-tk. center and Gaussian radius of puffs, and on the traillimages actually digitized using these programs. 1, PROGRAMMING CHANGES Program Improvements Problems Protracted work with existing trail digitizing programs used at Information Design's Pictorial Analysis Facility showed two deficiencies: 1. A serious problem was operator fatigue, since fatigue and resultant stress can detiorate the data reduction process. Fatigue was thought to be due to the requirement for carefully monitoring of a cathode ray tube (CRT) presentation of alphanumeric and pictorial information. Hardware and software considerations yielded a maximum flicker rate of about four frames per second. 2. Another problem was the rigid structuring of the program. The film analysis was performed as a series of operations which were executed primarily in a predetermined order. This implied that the operator was not free to perform ancillary tests or to check judgments except in prespecified ways. Solutions The operator fatigue problem was attacked primarily by making hardware changes. A CRT Terminal with integral memory and a programmable cursor capability was made the device on which.1 all alphanumeric information was displayed. Because this is a flicker-free display, much more information is now available to the operator at a glance. Messages indicationg cursor x and y position, last recorded position, tracking limits, filter settings, and available program function instructions are now available at all times. Pictorial information is now supplied by a CRT controller which incorporates a bit-oriented inatruction set. This makes it possible to display eight picture elements in the time formerly required to display one pixel. 2

6 The program has now been restructured -in such a way that any subroutine can be called from any other subroutine. This allows the operator greater flexibility to make cross-checks, detailed examination of scanning areas, and so forth. Any errors he might make, such as scanning an area that doesn't include the entire smoke trail, can now be easily remedied by re-performing the required subroutine. Each subroutine presents its available set of function options on the alphanumeric CRT, and the operator selects among these with sixteen microswitch pushbuttons. Results The only objective measure of the effect of these improvements would be a comparison of data reduction accuracy both before and after the changes. This comparison is not available. Subjectively, however, the results are dramatic. The operator now reports very little fatigue. The data reduction task has become much less stressful because error checks can easily be made, and ambiguous areas of trail pictures can be scanned at higher than normal magnification for more detailed analysis. One non-programming consideration which has proved very useful is the preparation by AFGL of photographic prints of each of the films. Thses have proved useful during the course of making decisions about trail ambiguities in the areas of trail self-intersections. Program additions SCentering Requirements The repertoire of available subroutines in the analysis program has been expanded in order to accommodate smoke puff images. These images are diffuse in character and of an approximately elliptical shape. Their density cross-sections should be approximately Gaussian. The position of the center of the puff is of importance for determining wind velocities, while the Gaussian radius is related to diffusion characteristics. Shape handling The diffuse, elliptical shapes have been accommodated by allowing the operator to scan rectangular areas of any desired aspect ratio. He controls the position, height, and width of a rectangular box displayed superimposed in the image of the entire frame. For each puff scanned, the box is adjusted to completely enclose the desired puff. Detrnxinig the center of puffs has been accomplished by defining the center to be the center-of-mass of the puff image. By "mass", of sourse, we mean the optical density of the image elements. To exclude neighboring images such as foreground objects, dust specks, streaks, adjacent puffs, etc. from the centroid computations, a medial axis transform (MAT) and inverse MAT have been implemented. These techniques allow the operator to *3

7 construct a mask which completely encloses large images, while totally excluding smaller images. See the Bibliography for references to a more detailed description of the MAT technique. After using the MAT to isolate the puff, minimum optical density is found, assumed to be a background level, and is therefore subtracted from all remaining MAT-masked densities. These densities are now used to mathematically wight the corresponding x, y coordinates of each point within the masked area. The weighted average of the mask defines the centroid. Limited experiments using a typical puff film have shown that even, very large differences in MAT mask size yield small defferences in resultant calculated centroid. Consistency on the order of plus-orminus 50 to 100 microns has been observed for puffs of several millimeters image size. We may infer that the output consistency despite input variability indicates a high degree of accuracy. Gaussian radius After the operator has determined the center of a puff, a major axis and a perpendicular minor axis are displayed superimposed on the puff image. Digital knobs can then be rotated to align these axes' with the puff orientation as judged by the operator. As the axes are rotated, the program calculates the approximate Gaussian radius for both axes and displays these on the alphanumeric CRT, Simultaneously, axis density cross-sections are displayed pictorially. The radius is found by first determining the total area under the cross-section profile, minus background density, along the axis. Then, starting at the center, successive points on both sides of the center are added until the integrated area is I- 1 /e times the total area. This is the Gaussian diameter, and is divided by two to yield the radius. Although this is an inelegant method, it has the advantage of being computationally rapid and thus allows real-time display of the results. Care has been taken to maintain spatial relationships properly, despite the differing distances between rectilinear raster elements during axis rotation. PHOTOGRAPHIC ANALYSIS For all events, at all sites, films were aligned on the scanner so that a line connecting a pair of points chosen on fiducials on opposite sides of the frame would be orthogonal with the axis of the scanner drum to a tolerance of 100 microns over the width of the frame. Films processed and digitized for smaoke trail triangulations are tabulated in the follwing table. Resolution on each frame was variable at the operator's discretion between 25 and 50 microns raster. 4

8 Name Frames Processed Project Paradise Aeolus 136 Winter Anomaly Program 19 July 1976 Stratospheric Trail 14 jand Trails, puffs, and starcals were sometimes taken from single films, all data were submitted in the form of magnetic tapes in CDC-6600 compatible format. 15

9 BIBLIOGRAPHY I. Quesada, A. F., "Application of Vector and Matrix Methods to Triangulation of Chemical Releases in the Upper Atmosphere", AFCRL , April NASA Technical Note N , October 1961, Langley AFB, Va. 3. Justus, C. G., "A Method Employing Star Backgounds for Improving the Accuracy of the Location of Clouds or Objects in Space fro Data Recorded on Film", A Thesis, Georgia Insitute of Technology, June Quesada, A.F. and Corbin, V. L., "Vector Detrmination of Camera Orientation from Stellar Photography", Air Force Cambridge Research Labs, Bedford, MA 5. Quesada, A. F., "Focal Length Correction for Triangulation Cameras", AFCRL, Bedford, Mass. 6. Wilson, G. D., Philbrick, 0., Michaels, S. B., and Morris, J. S., "Pictorial Analysis of Chemical Smoke Trails to Characterize a Wind Field in the KM Altitude Range", AFCRL-TR , AFCRL, Bedford, Mass. 7. Philbrick, 0., "Shape Recognition with the Medial Axis Transform", in Pictorial Pattern Recognition, Thompson Book Company, 1968, Washington, D. C. 8. Philbrick, 0., "SCRIPT, an Executive Program for a Reasearch Computer", AFCRL , November 1969, AFCRL, Bedford, Mass. 61